Now that Godot 4.0 is out and we have all returned from GDC we can finally take the time to solidify our release management plan for the 4.x releases (starting with 4.1). In our earlier blog post on the matter, we mentioned that we would like to shift to smaller, more regular releases.
During the 3.x development cycle we noticed two concerning trends:
- Contributors were crunching shortly before planned releases to add big features to those releases as they expected the next release would be a long ways away. This created a “now or never” mentality that led to us optimistically merging PRs that were not ready.
- Our “feature freeze” periods ended up being quite long as we often had a lot of big features crammed in quickly towards the end, and we continued to merge risky changes during our feature freeze period.
Our hope is to avoid both of these issues with our new approach to release management. We want to decrease the stress felt by both contributors and maintainers as we approach releases. We also want to be more cautious about merging big changes.
To address these concerns our plan is to shorten the release cycle and provide clear dates in advance. At the same time, we want to organize releases into two phases: the feature merging phase and the bug fixing phase. At the conclusion of the bug fixing phase we will release a new minor version and repeat the process.
Feature Merging Phase
The current plan is for the feature merging phase to be approximately 3 months long. For 4.1, this would mean it will cover March, April, and May 2023. During the feature merging phase we will gladly merge all pull requests that are ready to be merged and are approved by the relevant teams. This includes new features, regular bug fixes, and riskier bug fixes that we wouldn’t merge during the bug fixing phase.
Our goal is to reduce the pressure to get big changes merged “in time for the next release”. Since the next release will always be only a few months away, we can safely defer merging until the following release to give contributors time to properly implement and test new features. We hope this will enable us to have significantly shorter bug fixing phases.
For reference, Godot 4.0 had a 6 month “feature freeze” where we tried to avoid merging new features or especially risky bugs. This period was difficult for everyone and we would like to avoid such long periods of feature freeze in the future.
During the feature merging phase we will also cherry-pick safe bug fixes back to the current stable release (e.g. 4.0.x, 4.1.x, etc.).
We plan on releasing regular “dev snapshot” builds during the feature merging phase to assist users in testing and contributors in bisecting regressions. We encourage users to test out these dev snapshots on non-production projects to give us early feedback on new features. Similar to the 4.0-alphas, we expect to provide dev snapshots every 2 weeks or so with the frequency increasing as we approach the bug fixing phase.
Bug Fixing Phase
The current plan is for the bug fixing phase to be approximately 1 month long. For 4.1, we plan for the bug fixing phase to cover the month of June with our anticipated release of 4.1 being the end of June 2023.
We expect this time period will need tweaking as we see how much dedicated bug fixing time we need to stabilize after 3 months of feature merging.
To contrast this with our 3.x workflow, we tended to release minor versions once per year or so. Each release came with a feature freeze period with an indeterminate length, but usually several months.
During the bug fixing phase we plan on releasing weekly beta builds and eventually RC builds as we did in the latter part of 4.0’s development. Testing of beta builds will be highly encouraged as we want to spot and fix regressions quickly.
If the above goes well and we release Godot 4.1 by the end of June we will target the end of October for releasing 4.2. Naturally, if the release of 4.1 is delayed, our estimate for 4.2 will be delayed as well as we will still ensure we maintain our 3 month feature merging and 1 month bug fixing phases.
Since this is a new process, don’t be surprised if we make adjustments to the plan as we go along. We are still trying to figure out what sort of process works best with our contributors and our users. On that note, we appreciate feedback so please make sure to share your pain points so we can make this release process work for everyone.
Long Term Support of Stable Releases
If we follow the above schedule, we will be making around 3 minor releases per year. We expect that it will quickly become difficult to cherry-pick safe bug fixes back to all previous minor releases. In the Godot 3.x cycle we restricted cherry-picking to only the latest minor release and we only updated older releases when needed to fix critical issues.
We intend to minimize breaking changes in the 4.x release cycle, ideally there will be little reason to remain on older minor releases. However, we still want to ensure that users on older releases benefit from compatible bug fixes for as long as possible.
Right now we are unsure how many versions we can maintain in tandem. For now, we are committed to providing 4.0.x releases as we work on 4.1. Once 4.1 releases and our development focus shifts to 4.2, we will likely continue providing 4.0.x and 4.1.x releases with safe bug fixes. Once 4.2 releases we will be re-evaluating how many version back we can go with bug fix releases. At that point, development of 4.0.x may stop.
As always, we welcome feedback from the community and will be keeping a close eye on how many projects choose to remain on 4.0.x. If many projects choose to remain on 4.0.x, then that will influence how long we continue to provide 4.0.x releases.
We have just released the first beta for Godot 3.6 and we plan on releasing the stable version in the next few months. As the main focus of contributors is on Godot 4 and beyond, the development process for Godot 3 is becoming slower. As such, we don’t have an exact release date in mind, but we will be continuing to merge pull requests into the 3.x branch and will release 3.6 when it is ready.
Once we release 3.6, we don’t intend to make another minor release in the 3.x series. We will continue releasing 3.6.x bug fix updates for as long as we receive contributions to 3.6.
We have not started planning for Godot 5.0 yet. Our hope is to continue developing Godot 4.x for many years. Instead of planning for breaking changes we will work diligently to add new features without breaking compatibility. We believe the foundation we laid with 4.0 will make this much more tolerable than it was previously.
Godot is a non-profit, open source game engine developed by hundreds of contributors on their free time, and a handful of part or full-time developers hired thanks to donations from the Godot community. A big thank you to everyone who has contributed their time or financial support to the project!